The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses pnpx to fetch and execute the nopeek utility from the NPM registry. This tool is the core component providing the secret redaction and environment management functionality described in the skill.
[COMMAND_EXECUTION]: Executes shell commands to initialize cloud CLI profiles (such as AWS, GCP, and Azure), load secrets from local environment files, and audit directory structures for exposed credentials.
[PERSISTENCE]: Employs shell hooks (session-load.sh and redact-output.sh) that automatically trigger during session start and command execution to provide continuous secret protection across agent interactions.
[INDIRECT_PROMPT_INJECTION]: The skill handles data from potentially untrusted sources by scanning environment files and wrapping cloud CLI outputs to identify secrets.
Ingestion points: Reads local .env files and captures output from various third-party cloud tools (e.g., aws, kubectl) as specified in the redaction hook documentation.
Boundary markers: The tool implements its own boundary logic, using regex-based redaction to prevent secret patterns from being exposed to the agent context.
Capability inventory: Performs file system reads, environment variable injection via CLAUDE_ENV_FILE, and shell command wrapping.
Sanitization: Designed to sanitize tool outputs by redacting bearer tokens, private key headers, and database connection strings before they are returned to the agent.

nopeek